Corresponding author: Department of Psychology, Latino Alcohol and Health Disparities Research and Training Center, University of Texas at El Paso, 500 West University Avenue, El Paso, TX 79968, USA. E-mail: ude.petu.srenim@2sdrahcirkd
Received 2020 Jan 14; Revisions requested 2020 Mar 2; Revised 2020 Mar 2; Accepted 2020 Mar 2.Copyright © The Author(s) 2020. Medical Council on Alcohol and Oxford University Press. All rights reserved.
This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
This study assessed the factor, concurrent and predictive validity of the revised Readiness to Change Questionnaire [Treatment Version] (RCQ[TV]) among non-treatment-seeking individuals.
Non-treatment-seeking patients (Mage = 34.8, SD = 12.4) who screened positive for alcohol misuse were recruited from three urban Level I Trauma Centers and completed the RCQ[TV] (Heather et al. [(1999) Development of a treatment version of the Readiness to Change Questionnaire. Addict Res7, 63–83]).
A confirmatory factor analysis supported the three-factor structure of the RCQ[TV]. Observed scores for precontemplation, contemplation and action demonstrated concurrent validity, as they were correlated with drinking and alcohol-related problems prior to baseline assessment. Finally, RCQ[TV] scores at baseline added to the predictability of an alcohol consumption composite score at a 3-month follow-up after controlling for baseline alcohol consumption and randomization to treatment arm.
The results of the present study suggest that the RCQ[TV] has desirable psychometric properties and supports the use of the RCQ[TV] among non-treatment-seeking patients with alcohol misuse.
Short summary Although designed for treatment-seeking populations, the revised Readiness to Change Questionnaire [Treatment Version] (RCQ[TV]) for assessing the stages of change for reducing/quitting drinking may also be appropriate for use with non-treatment-seeking populations. The present study found that the RCQ[TV] had desirable psychometric properties among non-treatment-seeking individuals.
The stages of change (SoC) (Prochaska et al., 1992) is a common conceptualization of motivation to change alcohol use (Krebs et al., 2018). SoC describes the process of behavior change as progressing (often non-linearly) through five stages including precontemplation, contemplation, preparation, action and maintenance. These five stages represent varying degrees of readiness to change (RTC). SoC/RTC are proposed to be important constructs in the context of motivational interviewing (DiClemente and Velasquez, 2002), including brief motivational intervention (BMI) for alcohol misuse, a practical application of motivational interviewing delivered in opportunistic settings (e.g. emergency departments; Field et al., 2005).
Specifically, pretreatment SoC/RTC can be used to tailor BMI (Rollnick et al., 1992a), may affect alcohol outcomes above and beyond the influence of BMI (e.g. Bertholet et al., 2009) and may influence response to BMI (e.g. Barnett et al., 2010). Given the above, valid measures of SoC/RTC for use in the context of BMIs for alcohol misuse in opportunistic settings are needed. The present analyses test the validity of the Readiness to Change Questionnaire [Treatment Version] (Heather et al., 1999) among a sample of non-treatment-seeking trauma patients who received a brief intervention for alcohol misuse.
The Readiness to Change Questionnaire (RCQ) was initially developed to assess SoC/RTC for reducing alcohol use among non-treatment-seeking patients who are identified for alcohol misuse in medical settings (Rollnick et al., 1992b). The RCQ was developed by adapting items from the University of Rhode Island Change Assessment Scale (URICA) (Rollnick et al., 1992b). Twenty-eight items were assigned to one of four stages: precontemplation, contemplation, action or maintenance. Twenty items (5 items per stage) were retained for the initial validation study conducted among medical patients who screened positive for alcohol misuse (Rollnick et al., 1992b). Items reflecting the maintenance stage were eliminated due to inconsistency in responses.
Based on the results of an exploratory principal components analysis, an item from each stage was eliminated (3 items in all), resulting in 12 items in total (4 items per factor). The 12-item scale was then subjected to a principal component analysis that supported the adequacy of a three-factor structure. The 12-item scale demonstrated adequate concurrent validity with cartoons depicting each of the SoC and with alcohol-related questions included in a screening measure (Rollnick et al., 1992b). In a follow-up study, Heather et al. (1993) found support for the predictive validity of the RCQ such that non-treatment-seeking patients with alcohol misuse assigned to the action SoC demonstrated greater reductions in alcohol use at follow-up than those assigned to the precontemplation and contemplation SoC.
Heather et al. (1999) later adapted the RCQ to be used among treatment-seeking individuals, referred to as the Readiness to Change Questionnaire [Treatment Version] (RCQ[TV]). A major revision to the RCQ[TV] was that the instructions were adapted to reflect SoC/RTC for quitting drinking in addition to reducing drinking. In the initial development study of the RCQ[TV], the original set of 12 items from the RCQ was retained, while adding items from related scales (e.g. URICA and Stages of Change Readiness and Treatment Eagerness Scale; Miller and Tonigan, 1996). The authors also included items representing the preparation and maintenance stages in the RCQ[TV].
After several iterations of principal components analyses, a three-component structure that included 15 items representing the precontemplation, contemplation and action SoC was deemed as the most adequate (Heather et al., 1999). Concurrent validity was observed and internal consistency was mostly adequate. In a later study (Heather and Hönekopp, 2008), the authors tested the three-factor structure of the 15-item RCQ[TV] via confirmatory factor analysis (CFA). The CFA results indicated somewhat poor model fit indices, and thus, the authors examined potentially inadequate items. As such, the item with the lowest loading on each of the scales was eliminated, resulting in a revised 12-item scale with improved model fit indices (Heather and Hönekopp, 2008). Adequate subscale reliabilities (precontemplation α = 0.66–0.77; contemplation α = 0.66–0.82; action α = 0.85–0.88) were found, with higher reliability observed for 3- and 12-month follow-ups. Additionally, participants were allocated to either pre-action (i.e. precontemplation and contemplation) or action SoC, and those allocated to the action SoC demonstrated greater reductions in alcohol use 3 and 12 months after treatment.
Despite the adequacy and convenience of the revised RCQ[TV], few published studies have included it as an assessment tool, and no studies have assessed its factor structure aside from the original validation study (Heather and Hönekopp, 2008). We also know of no study that has replicated the revisions to the RCQ[TV] (i.e. 15-item to 12-item version). Although the RCQ[TV] is intended for use with treatment-seeking populations, the adaptations that were made may also be improvements for use among non-treatment-seeking populations. The RCQ[TV] reflects SoC for both reducing and quitting drinking and thus may better capture the diverse goals of non-treatment-seeking populations. Given the differences between treatment-seeking and non-treatment-seeking people with alcohol use disorder (e.g. Ray et al., 2017; Rohn et al., 2017; Lee et al., 2019), it is unclear whether the RCQ[TV] will demonstrate validity among non-treatment-seeking populations. Further, researchers have traditionally scored the RCQ and RCQ[TV] using stage allocation. However, SoC has been criticized conceptually for stage allocation (e.g. Sutton, 2001; West, 2005), and the use of continuous scores has statistical benefits such as sensitivity to change (Carey et al., 1999; Rucker et al., 2015). Thus, psychometric testing of the continuous RCQ[TV] subscale scores is also warranted.
For the present study, we used baseline and 3-month follow-up data from a randomized controlled trial of non-treatment-seeking patients who screened positive for alcohol misuse and subsequently received a brief intervention. First, we assessed the factor structure of the RCQ[TV], which included comparisons of the 15- and 12-item versions. Based on previous studies (Heather et al., 1999; Heather and Hönekopp, 2008), we hypothesized that a three-factor structure of the 12-item RCQ[TV] would provide the best fit to the data. Second, we assessed associations between RCQ[TV] observed scores and alcohol use and related consequences for the prior 3 months at baseline. We hypothesized that higher precontemplation would be associated with less alcohol use and fewer alcohol-related problems 3 months prior, whereas higher contemplation and action would be associated with more alcohol use and greater alcohol-related problems 3 months prior. These hypotheses were based on findings that patients with more severe alcohol misuse are more motivated to change their drinking (Kahler, 2001; Williams et al., 2006; Krenek et al., 2011; Reed et al., 2019). We also examined the predictive validity of the RCQ[TV] subscales beyond the effects of intervention as well as interaction effects with the intervention. Consistent with Heather and Hönekopp (2008), we hypothesized that higher precontemplation and contemplation scores would predict increased alcohol consumption at 3-month follow-up and higher action scores would predict decreased alcohol consumption at 3-month follow-up.
Secondary data analyses were conducted using baseline and 3-month follow-up data from a larger multisite randomized controlled trial of brief intervention in the trauma care setting (Field et al., 2014). Readers are referred to Field et al. (2014) for further information about the primary aims, methods and findings of the trial. In brief, Field et al. found that BMI with a telephone booster resulted in greater reductions in alcohol use at 3-, 6- and 12-month follow-ups compared to brief advice and BMI without a telephone booster. Ethical approval for the trial was obtained from each participating institution and a certificate of confidentiality was obtained from the U.S. Department of Health and Human Services.
Participants were 596 injured patients (see Table 1 for baseline characteristics of the sample) recruited from three urban Level I trauma centers in Texas from October 2007 to December 2010.
Baseline characteristics of the sample
| Variable | f | % |
|---|---|---|
| Gender | ||
| Male | 456 | 76.5 |
| Female | 140 | 23.5 |
| Ethnicity | ||
| Non-Hispanic White | 237 | 39.8 |
| Black | 158 | 26.5 |
| Hispanic | 172 | 28.9 |
| Other | 29 | 9.8 |
| Education | ||
| Less than high school | 62 | 10.4 |
| High school degree | 199 | 33.4 |
| Vocational training | 241 | 40.4 |
| College degree | 57 | 9.6 |
| Missing | 37 | 6.2 |
| Marital status | ||
| Never married | 277 | 46.6 |
| Married/living with someone | 156 | 26.2 |
| Separated/divorced/widowed | 162 | 27.2 |
| Employment status | ||
| Full-time job | 259 | 43.5 |
| Part-time job | 86 | 14.4 |
| Unemployed | 177 | 29.7 |
| Not in labor force | 30 | 5.0 |
| In jail, prison, or detention | 1 | 0.2 |
| Too disabled for work | 33 | 5.5 |
| Other situation | 10 | 1.7 |
| M (SD) | Range | |
| Age | 34.79 (12.39) | 18–81 |
| Total family income | 42107.81 (62,292.39) | 0–1,000,000 |
| % days abstinent | 66.89 (30.39) | 0–99 |
| % days heavy drinking | 22.03 (26.75) | 0–100 |
| Total no. of drinks | 236.20 (388.25) | 1–4204 |
| Average no. of drinks | 18.22 (29.81) | 0–316 |
| Maximum no. of drinks | 13.43 (11.12) | 1–78 |
| DDD | 2.65 (4.37) | 0–47 |
| SIP +6 | 11.57 (12.52) | 0–58 |
Note. Alcohol use was assessed during the past 3 months. DDD, number of drinks per drinking day.
Patients treated for unintentional or intentional injuries, violence-related injuries and assault-related injuries were screened for study participation. Patients who were intoxicated or presented with a Glasgow Coma Scale < 14 were screened once medically stable. Further, patients had to have had orientation to person, place and time and adequate recall of recent and remote events, assessed via the Mini-Mental Status Exam, to be screened. Patients screened positive based on at least one criterion: (a) positive blood alcohol concentration (BAC >0.01) at the time of admission, (b) self-report of drinking 6 hours prior to injury or (c) gender-specific cutoff scores (≥3 for women; ≥4 for men) on the consumption questions from the Alcohol Use Disorders Identification Test (Bradley et al., 2007). See Field et al. (2014) for exclusion criteria.
Patients who met the eligibility criteria and consented to participate completed a baseline assessment. Following the baseline assessment, participants were randomly assigned to one of three conditions: brief advice (BA; n = 200), a minimal intervention condition; BMI (n = 203) (for a description of BMI in the trauma care setting, see Field et al., 2005) or BMI plus a telephone booster using personalized drinking feedback that occurred about 1 month after the initial intervention (BMI + B; n = 193). Follow-up assessments were conducted 3, 6 and 12 months post-intervention. Research staff conducting the baseline and follow-up assessments were blind to intervention condition. Measures relevant to the present study are described below.
The RCQ[TV], which was previously described in further detail, was used to assess precontemplation, contemplation and action at baseline and 3-month follow-up. Items are scored using a Likert-type scale ranging from −2 (strongly disagree) to +2 (strongly agree). The precontemplation (PC), contemplation (C) and action (A) scores were computed by summing across the respective items.
Alcohol use during the past 3 months (for both baseline and 3-month follow-up) was assessed using the Timeline Follow-Back (TLFB; Sobell and Sobell, 1992; Sobell et al., 1979). We used the following six alcohol use variables derived from the TLFB: percent days abstinent, percent days heavy drinking (>4 drinks for men or >3 drinks for women during a single drinking episode; Dawson et al., 2005), total number of drinks consumed, average number of drinks consumed per week, maximum number of drinks consumed on one occasion and average number of drinks consumed per drinking day. Total number of drinks consumed, average number of drinks per week, maximum number of drinks consumed on one occasion and average number of drinks per drinking day were log-transformed due to non-normality.
The 21-item Short Inventory of Problems +6 (SIP +6; Soderstrom et al., 2007) was used to assess alcohol-related consequences (e.g. ‘I have failed to do what is expected of me because of my drinking’) experienced over the past 3 months at baseline. However, the SIP +6 was not included in the 3-month follow-up assessment. The estimate of Cronbach’s alpha for the SIP +6 in the present study was high (α = 0.94).
A CFA for the hypothesized three-factor structure of the 12-item version of RCQ[TV] was estimated in Mplus 8 (Muthén and Muthén, 1998–2017) using MLR estimation, which is a maximum likelihood estimator with robust standard errors that accounts for missing data. Item loadings were estimated and the factor variances were set to 1. Covariances between the latent factors were estimated. As shown in Table 2 , the model yielded adequate fit to the data (based on joint global model fit criteria proposed by Hu and Bentler, 1999). Table 3 presents the standardized factor loadings and standard errors as well as the item descriptives. Further, the correlation between the precontemplation and contemplation factors was −0.836. The correlation between the precontemplation and action factors was −0.499. Finally, the correlation between the contemplation and action factors was 0.604 (all Ps ≤ 0.01).
Summary of the global fit indices for the hypothesized and alternative models of the RCQ[TV]
| Model | SB Χ 2 | df | CFI | RMSEA (90% CI) | SRMR | AIC |
|---|---|---|---|---|---|---|
| Hypothesized model (12 items) | ||||||
| Three factors | 220.86 | 51 | 0.922 | 0.075 (0.065, 0.085) | 0.065 | 20001.89 |
| Alternative models (12 items) | ||||||
| Two factors | 266.32 | 53 | 0.902 | 0.082 (0.073, 0.092) | 0.066 | 20058.81 |
| One factor | 653.31 | 54 | 0.724 | 0.137 (0.127, 0.146) | 0.095 | 20543.25 |
| Alternative models (15 items) | ||||||
| Three factors | 445.96 | 87 | 0.874 | 0.083 (0.076, 0.091) | 0.073 | 24794.36 |
| Two factors | 506.47 | 89 | 0.853 | 0.089 (0.081, 0.096) | 0.076 | 24871.08 |
| One factor | 820.40 | 90 | 0.744 | 0.117 (0.100, 0.124) | 0.086 | 25264.32 |
Note. SB χ 2 = Satorra–Bentler scaled χ 2 ; CFI, comparative fit index; RSMSEA, root mean square error of approximation; CI, confidence interval; SRMR, standardized root mean square residual.
Standardized factor loadings (standard errors) and unique variances (standard errors) and item descriptives of the three-factor RCQ[TV]
| Item | λ (SE) | Ψ 2 (SE) | M (SD) |
|---|---|---|---|
| PC | |||
| Item 1 | 0.673 (0.037) | 0.547 (0.050) | 0.06 (1.31) |
| Item 4 | 0.786 (0.036) | 0.383 (0.056) | 0.17 (1.26) |
| Item 7 | 0.332 (0.052) | 0.890 (0.034) | 0.39 (1.08) |
| Item 13 | 0.671 (0.038) | 0.550 (0.051) | −0.37 (1.17) |
| C | |||
| Item 2 | 0.694 (0.031) | 0.518 (0.044) | 0.08 (1.25) |
| Item 5 | 0.689 (0.033) | 0.526 (0.045) | 0.49 (1.19) |
| Item 8 | 0.795 (0.024) | 0.368 (0.038) | −0.13 (1.26) |
| Item 14 | 0.780 (0.026) | 0.392 (0.040) | −0.22 (1.23) |
| A | |||
| Item 6 | 0.560 (0.039) | 0.686 (0.043) | 0.29 (1.13) |
| Item 9 | 0.755 (0.027) | 0.430 (0.041) | 0.58 (1.13) |
| Item 12 | 0.853 (0.023) | 0.273 (0.038) | 0.36 (1.17) |
| Item 15 | 0.803 (0.026) | 0.356 (0.042) | 0.16 (1.19) |
Note. PC, precontemplation subscale of the RCQ[TV]; C, contemplation subscale of the RCQ[TV]; A, action subscale of the RCQ[TV].
We then tested alternative models to address potential bias toward confirming the three-factor structure of the 12-item RCQ[TV] that was hypothesized (Brown, 2015).
Given the large correlation between the precontemplation and contemplation factors for the three-factor structure, the first alternative model that we tested was a two-factor structure where the precontemplation (reverse scored) and contemplation items loaded onto a common factor. Previous studies (e.g. Hannöver et al., 2002) have found support for a two-factor structure of the original RCQ in which precontemplation and contemplation load onto a common factor and other studies (e.g. Heather and Hönekopp, 2008) have combined precontemplation and contemplation groups to form a ‘pre-action’ group. The second alternative model that we tested was a one-factor structure. Previous studies (e.g. Budd and Rollnick, 1996) have used the subscale scores of the RCQ[TV] to compute a total RTC score ([C + A] − PC). Thus, it may be that the items of the RCQ[TV] represent a single, continuous RTC construct.
We conducted CFAs for the two alternative models for both the 12- and 15-item versions of the RCQ[TV] as well as a three-factor structure for the 15-item version to provide a comprehensive comparison of the 12- and 15-item versions. We compared the global fit indices and Akaike Information Criterion (AIC) values, with the lowest AIC being the preferred model (Brown, 2015), of the alternative models to the hypothesized model. As shown in Table 2 , both the global fit indices and AICs indicated that the three-factor structure of the12-item version of the RCQ[TV] was the preferred model.
We then estimated the reliability of the three subscales at baseline using McDonald’s coefficient omega (McDonald, 1999). Reliability of the items forming the PC subscale yielded a McDonald’s omega of 0.68, reliability of the items forming the C subscale had a McDonald’s omega of 0.83 and reliability of the items forming the A subscale had a McDonald’s omega of 0.84. We also estimated the test–retest reliability of each subscale by correlating the subscale score at baseline with its respective subscale score at 3-month follow-up among participants assigned to the BA condition, given that the BMI conditions were designed to increase motivation to reduce or quit drinking. The correlation between the PC subscale scores at baseline and 3-month follow-up was r = 0.48. The correlation between the C subscale scores at baseline and 3-month follow-up was r = 0.55. The correlation between the A subscale scores at baseline and 3-month follow-up was r = 0.29 (all Ps < 0.001).
To test the concurrent validity of the RCQ[TV], we assessed the correlations of the RCQ[TV] subscales with indices of alcohol use and SIP +6 scores for the past 3 months. These correlations are reported in Table 4 . As shown in Table 4 , PC scores were positively correlated with percent days abstinent and negatively correlated with each of the other indices of prior drinking. C and A scores were negatively correlated with percent days abstinent and positively correlated with each of the other indices of prior drinking. Additionally, PC scores were negatively correlated with SIP +6 scores, while C and A scores were positively correlated with SIP + scores.
Bivariate correlations between the RCQ[TV] subscales and drinking and alcohol-related problems prior to baseline
| PC | C | A | |
|---|---|---|---|
| Drinking/alcohol problems (baseline) | |||
| % days abstinent | 0.204 ** | −0.336 ** | −0.094 * |
| % days heavy drinking | −0.271 ** | 0.409 ** | 0.141 ** |
| Log-transformed total no. of drinks | −0.283 ** | 0.491 ** | 0.167 ** |
| Log-transformed average no. of drinks | −0.300 ** | 0.490 ** | 0.160 ** |
| Log-transformed maximum no. of drinks | −0.254 ** | 0.418 ** | 0.172 ** |
| Log-transformed DDD | −0.271 ** | 0.418 ** | 0.175 ** |
| SIP +6 | −0.591 ** | 0.655 ** | 0.348 ** |
Note. Alcohol use was assessed during the past 3 months.
Given that the correlations of the C and A scores with prior drinking were in the same direction, we compared the magnitude of the correlation coefficients using Lee and Preacher’s (2013) online utility for calculating the test of the difference between two dependent correlations with one variable in common. These tests revealed that the correlation coefficients for C scores with each of the indices of prior drinking were larger than those for A scores (all Ps < 0.05).
Finally, we tested the predictive validity of the RCQ[TV] in the context of the intervention by conducting a hierarchical multiple regression analysis. A composite alcohol use variable at 3-month follow-up was the outcome in the regression model. Rather than conducting separate regression analyses for each of the six alcohol use variables, we standardized and then averaged the log-transformed average number of drinks per week, maximum number of drinks consumed on one occasion and average number of drinks per drinking day variables to create a composite alcohol use variable. These three variables were chosen because the primary outcome analyses demonstrated that the intervention conditions had similar effects on the three variables at 3-month follow-up (Field et al., 2014). Further, correlations among these three variables were large in magnitude (r = 0.66–0.89, all Ps < 0.001), and others (e.g. Barnett et al., 2010) have used similar procedures for creating an alcohol use composite variable for brief intervention trials delivered in emergency departments.
The form of the regression was as follows: In Step 1, we entered the baseline alcohol use composite to model change in alcohol use from baseline to 3-month follow-up. In Step 2, we entered two dummy coded variables representing the effects of the BMI and BMI + B conditions as compared to BA. In Step 3, we entered the three RCQ[TV] subscales. In Step 4, we entered the six interaction terms between the two dummy coded variables representing the intervention conditions and mean-centered RCQ[TV] subscale scores. The results of the regression analyses are presented in Table 5 . As expected, baseline alcohol use predicted alcohol use 3 months later. After controlling for baseline alcohol use, individuals who received BMI + B had reduced alcohol consumption at 3-month follow-up, relative to individuals who received BA. Baseline scores on the RCQ[TV] furthered predicted alcohol use, such that higher C scores were associated with increased alcohol consumption at 3-month follow-up and higher A scores were associated with decreased alcohol consumption at 3-month follow-up. RCQ[TV] scores did not moderate the effect of the intervention.
Regression analysis predicting change in alcohol use from baseline to 3-month follow-up
| Variable | 95% CI | |||||
|---|---|---|---|---|---|---|
| ΔR 2 | B | LLCI | ULCI | β | P value | |
| Step 1 | 0.14 ** | |||||
| Alcohol use (baseline) | 0.39 | 0.30 | 0.47 | 0.37 | 0.000 | |
| Step 2 | 0.01 * | |||||
| BMI | −0.12 | −0.29 | 0.06 | −0.06 | 0.179 | |
| BMI + B | −0.26 | −0.46 | −0.05 | −0.12 | 0.014 | |
| Step 3 | 0.03 ** | |||||
| PC | −0.01 | −0.03 | 0.02 | −0.04 | 0.485 | |
| C | 0.03 | 0.00 | 0.06 | 0.14 | 0.043 | |
| A | −0.04 | −0.06 | −0.02 | −0.21 | 0.000 | |
| Step 4 | 0.01 | |||||
| BMI × PC | −0.01 | −0.07 | 0.05 | −0.03 | 0.720 | |
| BMI × C | 0.01 | −0.05 | 0.07 | 0.04 | 0.680 | |
| BMI × A | −0.04 | −0.09 | 0.01 | −0.12 | 0.100 | |
| BMI + B × PC | −0.04 | −0.11 | 0.02 | −0.09 | 0.171 | |
| BMI + B × C | −0.02 | −0.08 | 0.04 | −0.05 | 0.546 | |
| BMI + B × A | −0.03 | −0.09 | 0.02 | −0.08 | 0.252 | |
Note. Alcohol use was assessed during the past 3 months. LLCI = lower limit confidence interval. ULCI = upper limit confidence interval. BMI + B = Brief motivational intervention with telephone booster session.
The aim of the present study was to provide a test of the validity of the RCQ[TV] among a sample of non-treatment-seeking patients who screened positive for alcohol misuse. In support of our hypothesis, a three-factor structure of the revised RCQ[TV] (12 items) provided the best fit to the data. To our knowledge, this is the first replication of the factor structure and revisions to the RCQ[TV] among any sample. Importantly, these finding are consistent with previous studies that found support for a three-factor structure of the RCQ[TV] among treatment-seeking populations (Heather et al., 1999; Heather and Hönekopp, 2008). The findings confirm the factor structure in patients with alcohol misuse who are not seeking treatment.
Second, we tested the concurrent validity of the RCQ[TV] by examining baseline correlations of the RCQ[TV] subscales with indices of prior drinking and alcohol-related consequences. Higher PC scores were associated with less alcohol use and alcohol-related consequences during the past 3 months, whereas higher C and A scores were associated with greater prior consumption and consequences. Notably, the correlations were statistically different and of greater magnitude for contemplation compared to action. In other words, patients with greater alcohol misuse severity had greater RTC, which is consistent with previous studies among people who screened positive for alcohol misuse (Kahler, 2001; Williams et al., 2006; Reed et al., 2019) and/or expressed an interest in reducing their drinking (Krenek et al., 2011). One interpretation of these findings is that those with less severe alcohol misuse are less likely to perceive their drinking as problematic and therefore lack motivation to change. In contrast, those with more severe alcohol misuse are more likely to perceive their drinking as problematic and therefore be motivated to change. The comparatively larger correlations for contemplation compared to action seem to reflect the associated reductions in alcohol use and alcohol-related consequences resulting from taking steps toward reducing or quitting drinking. Given the above, the observed correlations between the RCQ[TV] subscales and prior drinking and alcohol-related problems support the concurrent validity of the RCQ[TV] for use with non-treatment-seeking populations.
Our final analyses tested the predictive validity of the RCQ[TV] in the context of the intervention. Both C and A scores predicted alcohol use above and beyond the effects of the intervention; however, PC scores did not. Higher C scores predicted increases in alcohol use from baseline to 3-month follow-up and higher A scores predicted decreases in alcohol use from baseline to 3-month follow-up, which is consistent with previous studies among both non-treatment-seeking (Bertholet et al., 2009) and treatment-seeking (Heather and Hönekopp, 2008) populations. These findings suggest that additional booster sessions may be needed for those who remain ambivalent about making changes (i.e. contemplation) following the brief intervention. We also examined whether the RCQ[TV] subscale scores influenced the effects of the intervention by testing interaction effects. While none of the interaction effects were statistically significant, support for RTC/SoC as a moderator of the effects of BMI delivered in medical settings is mixed (Walton et al., 2008; Saitz et al., 2009; Barnett et al., 2010). Overall, the findings of the present study support the validity of the RCQ[TV] for use among non-treatment-seeking patients identified for alcohol misuse.
A highly relevant implication of the present findings is the scoring the RCQ[TV] for use in future studies. Although overall RTC scores are often calculated (Budd and Rollnick, 1996; Forsberg et al., 2004; Borsari et al., 2009), the present findings caution against this practice, at least among non-treatment-seeking populations. Specifically, the differential associations among the latent factors in the CFA, the difference in magnitude of the baseline correlations for the C and A scores with prior drinking and alcohol-related problems and the opposite directions of the prospective associations for the C and A scores with drinking and alcohol-related problems all suggest against a higher-order factor structure. Thus, caution is warranted against using a total RCQ[TV] score ([C + A] − PC). Based on the current findings, the more robust approach may be to simultaneously examine the individual stages scores in any given statistical analysis.
There are several limitations of the present study that should be acknowledged. First, while the use of the 15-item RCQ[TV] afforded us the opportunity to replicate revisions made by Heather and Hönekopp (2008), it is unclear how responding to the additional three items in the 15-item version may have affected responses to the other 12 items that were subjected to concurrent and validity analyses. Second, it has been argued that SoC/RTC may be an important mechanism of change underlying the effectiveness of BMI (Apodaca and Longabaugh, 2009), but this was not explored in the present study. SoC/RTC as a mechanism of change has received little support (e.g. Stein et al., 2009), and thus we did not believe this to be an appropriate test of validity of the RCQ[TV]; this is being tested as a separate analysis of these data, however. Third, Field et al. (2014) detail other limitations of these data that may affect the generalizability of the findings, such as a moderate consent rate among patients who met the study’s eligibility criteria. Finally, the heterogeneity in alcohol misuse severity of the sample may also be a limitation as RTC reduce/quit drinking may be less applicable to those with less severe alcohol misuse. Future research should test the measurement invariance of the RCQ[TV] across these groups of drinkers based on alcohol misuse severity as these groups may interpret RTC differently.
The validation of the RCQ[TV] in this large sample of diverse (e.g. ethnicity, alcohol misuse severity) non-treatment-seeking patients constitutes a significant contribution to the literature as the use of the RCQ[TV] may be more appropriate than the RCQ because it reflects both harm reduction and abstinence as goals. The present study also highlights the importance of adequately assessing SoC/RTC in relation to alcohol use among non-treatment-seeking patients, which is given increasing attention because of the widespread application of brief intervention. The findings for SoC/RTC as both a moderator and mediator of BMI have been mixed, and the present study suggests that the way in which SoC/RTC measures are scored is important. That is, summing C and A scores may be especially problematic because this may obscure the unique baseline and prospective associations of these subscales with drinking and alcohol-related problems. Nevertheless, future studies are needed to further explore the use of the RCQ[TV] in this context, given the proposed importance of SoC/RTC to BMI. The results from the present study establish the construct, concurrent and predictive validity of the RCQ[TV] to robustly support its use in pursuit of these future aims.
This work was supported by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) [R01AA015439 to C.A.F.]. The contents of the manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the NIAAA. The NIAAA had no role in the study design; collection, analysis and interpretation of data; in the writing of the report; nor in the decision to submit the article for publication.
Articles from Alcohol and Alcoholism (Oxford, Oxfordshire) are provided here courtesy of Oxford University Press